Discourse Structure

Discourse structure research focuses on understanding how sentences and paragraphs connect to form coherent texts, aiming to model the underlying relationships and organization of information. Current research emphasizes leveraging large language models (LLMs) like GPT, along with graph neural networks (GNNs) and other deep learning architectures, to automate tasks like discourse parsing, relation classification, and summarization, often incorporating rhetorical structure theory (RST) and other frameworks. These advancements improve the efficiency and accuracy of analyzing complex texts, with applications ranging from automated content analysis in the social sciences to enhancing information retrieval and text generation systems. The development of new datasets and evaluation benchmarks is also a key focus, driving progress in cross-lingual and multimodal discourse analysis.